Seed Audio today announced support for Doubao Seed-Audio 1.0, the newly released multimodal audio generation model from ByteDance and Volcengine, inside its AI music creation workspace. This integration signals a shift in AI audio from isolated text-to-speech or single-track generation toward full-scene audio creation, where dialogue, emotion, accents, background music, ambience, and sound effects can be produced cohesively.
Doubao Seed-Audio 1.0, described in public launch coverage as a multimodal model that works with text and reference audio, is positioned for end-to-end audio creation rather than isolated clips. This distinction matters for creators working on projects like podcast trailers, short dramas, or game teasers, which require multiple audio elements to be synchronized. Unlike traditional text-to-speech models that focus on how words are spoken, Doubao Seed-Audio 1.0 addresses the broader sound of a scene, including voices, music, spatial texture, sound effects, and timing.
Seed Audio places this capability inside an agent-based environment designed to help creators move from first idea to usable audio. At the center is Seed Audio Agent, which translates plain-language goals—such as a cinematic game loop or a podcast intro—into clear music directions, selects creation or editing paths, and suggests follow-up actions. The platform includes tools for generating complete songs, instrumental tracks, hooks, and demos from text prompts via the AI Music Generator. Users can also access lyric and style assistance to refine vague ideas into structured instructions.
For workflows starting from existing material, Seed Audio supports uploads, reference tracks, and saved works. The AI Cover feature creates new vocal or style versions from source tracks, while Extend helps lengthen tracks for videos, podcasts, or game loops. Add Tracks enables accompaniment or vocal additions to partial recordings, and Mashup combines source ideas into new results. Replace Section allows targeted revision of weak parts, and Vocal Remover separates vocals and instrumentals for remixing or karaoke versions.
Seed Audio also includes discovery and library features. Through Explore, users can browse public tracks to understand different prompts and genres. My Works manages previous generations for further editing or agent-guided revision. The platform is especially useful for creators needing music that fits specific formats, such as background tracks with room for narration, short intros with clean endings, or loopable instrumentals.
"Doubao Seed-Audio 1.0 shows where AI audio is heading, toward richer, more contextual creation," said a Seed Audio spokesperson. "Our goal is to make that capability useful inside a real creator workflow. Creators do not just need a model response. They need a way to draft, refine, reuse, and finish audio assets."
Seed Audio is available now at https://seedaudio.ai. New users can start with Seed Audio Agent, test Doubao Seed-Audio 1.0-supported workflows, generate sample tracks, and use the platform's tools. For creators needing visual assets, i2v.ai offers AI image and video generation that pairs with Seed Audio for short videos, social posts, and campaign assets.
